A 5-day course covering Google's Agent Development Kit (ADK) and advanced agent architecture patterns. You can see the course here
Also contains a simple nixenv and direnv setup for environment management
- Agent Fundamentals: Model, tools, and orchestration layers
- Multi-Agent Systems: Sequential, parallel, and loop agents
- Agent Tools & Integration: Custom tools, MCP, and pausable workflows
- Context Engineering: Memory management and session handling
- Agent Security & Interoperability: Guardrails and A2A protocols
Built using Google Agent Development Kit (ADK) with Python and Gemini models. Each day has it's associated code which you can find in the repo.
This repo contains my notes on the technical aspects as well as the whitepapers from the course (yes I read all of them)
day-1/day-1.md- Code notes: Multi-agent patterns and orchestrationday-1/introduction_to_agents.md- Whitepaper notes: Agent taxonomy and architecture fundamentals
day-2/day-2.md- Code notes: Agent tools, custom functions, and pausable workflowsday-2/agent-tools-mcp.md- Whitepaper notes: MCP integration and tool development
day-3/day-3.md- Code notes: Sessions, Memory and how to use themday-3/context_engineering_sessions_memory.md- Whitepaper notes: Context engineering, sessions, and memory systems
day-4/day-4.md- Code notes: How to asses agent quality with test casesday-4/agent-quality.md- Whitepaper notes: How to evaluate agent quality and principles for designing high quality agents
day-5/day-5.md- Code notes: MCP and deploymentday-5/prototype-to-production.md- Whitepaper notes: Moving local agent to production - best practices